基于属性依赖度的图像隐写分析算法  

Image Steganalysis Algorithm Based on Attribute Dependent Degree

作　　者：余文琼[1] 

机构地区：[1]三明学院数学与计算机科学系,福建三明365004

出　　处：《三明学院学报》2010年第6期506-509,537,共5页Journal of Sanming University

基　　金：福建省教育厅B类项目(JB10172);三明学院科研基金项目计划(2009B0901/G);三明学院教改科研项目(ZL0701/SF)

摘　　要：针对文献[6]将粗糙集属性约简应用于信息隐藏盲检测中检测正确率有所下降的问题,提出了基于属性依赖度的图像隐写分析算法,该算法利用粗糙集理论属性依赖度提出决策表离散优化的措施,寻找一种提高整个决策表分类能力的办法,以达到提高检测正确率的目的。首先利用该算法对决策表进行优化,其次通过属性约简得到最小约简,最后采用支持向量机构造分类器,对Cox、Piva两种不同隐写术进行实验结果表明,使用该算法不仅检测正确率有较大提高,而且检测效率也有较大提高。The accuracy of hidden information blind detection in which rough set attribute reduction is adopted would drop to some extent in ref[6].Steganalysis algorithm based on decision attribute dependent degree is proposed.In this algorithm,improved measures for the discretization of the decision table are proposed to get the best decision ability of discretized decision table with the detection accuracy improved,according to the knowledge of decision attributes dependent degree in rough set theory.Experiments are taken to Cox,Piva as the following order： firstly,optimize the decision table by this algorithm;secondly,get the minimal feature after attributes reduction;thirdly,build the classifier by SVM.The experimental results shows that the detection accuracy and efficiency have been greatly improved by the help of the proposed algorithm.

关 键 词：隐写分析 检测正确率 属性依赖度 分类能力 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]